The final step in creating an AI agent is to set the strictness level. This determines how strictly the AI agent will follow the guidelines you’ve set.

There are 4 strictness levels:

  • Easygoing
  • Balanced
  • Moderate
  • Strict

The strictness level is less priority than the guidelines themselves. If you need to adjust your AI agent, adjusting the guidelines themselves is more important than adjusting the strictness level.

If you’re using Llama Guard we recommend to use the threshold setting in your project instead of tinkering with the strictness level. GPT models do not return a probability score for setting the threshold, so here the strictness level can be more useful.

Was this page helpful?